Intelligent evaluation method and system for optic nerve function based on visual evoked potential

Abstract

The invention relates to the technical field of ophthalmology informatization, in particular to an intelligent optic nerve function evaluation method and system based on visual evoked potential, and the method comprises the steps: obtaining an electroencephalogram signal, an eye movement track and an auditory reaction signal of a subject; obtaining visual stimulation parameters, wherein the visual stimulation parameters comprise stimulation frequency, stimulation intensity and stimulation duration; on the basis of the electroencephalogram signals, the eye movement tracks and the auditory reaction signals, signal preprocessing operation is executed, and preprocessed multi-modal physiological signals are obtained; extracting features by adopting an improved space-time convolutional neural network based on the preprocessed multi-modal physiological signals to obtain optic nerve function features; according to the optic nerve function features and the visual stimulation parameters, establishing an optic nerve conduction pathway model; evaluating the optic nerve function state based on the optic nerve conduction pathway model to obtain an optic nerve function evaluation result; generating an evaluation report containing the optic nerve function evaluation result; predicting the optic nerve function change trend, and generating a prediction report.

Description

Technical Field

[0001] The present invention relates to the field of ophthalmic information technology, and more specifically, to an intelligent evaluation method for optic nerve function based on visual evoked potential and a system thereof. Background Art

[0002] Accurate assessment of optic nerve function is crucial for the diagnosis and treatment of ophthalmic diseases. Traditional methods of optic nerve function assessment mainly rely on subjective tests such as visual acuity examination and visual field examination. Although these methods are simple and intuitive, they often fail to accurately reflect the actual functional status of the optic nerve. In recent years, the development of visual evoked potential (VEP) technology has provided an objective and quantitative method for optic nerve function assessment. VEP can reflect the integrity of optic nerve conduction function by recording the potential changes generated in the cerebral cortex after visual stimulation.

[0003] However, the existing VEP-based optic nerve function assessment methods still have some limitations. First, traditional VEP analysis mainly focuses on a few limited parameters such as the latency and amplitude of the P100 wave, which is difficult to fully reflect the complexity of optic nerve function. Secondly, the existing methods usually use simple filtering and averaging techniques to process VEP signals, which cannot effectively remove various interferences and noises, affecting the accuracy of the assessment. Furthermore, the current assessment methods lack the ability of individualization and dynamic adjustment, and are difficult to adapt to the characteristics and state changes of different subjects. In addition, the existing technology is also difficult to achieve accurate prediction of the trend of changes in optic nerve function, limiting its application in early disease prevention and long-term monitoring.

[0004] Finally, the current evaluation system often uses VEP as a single evaluation indicator, ignoring the supplementary information that other related physiological signals (such as eye movements, auditory responses, etc.) may provide, which to some extent limits the comprehensiveness and accuracy of the evaluation. These problems seriously restrict the clinical application of VEP technology in the evaluation of optic nerve function. Summary of the invention

[0005] The present invention aims to solve the above technical problems and provide a method and system for intelligent evaluation of optic nerve function based on visual evoked potential. The method achieves a comprehensive, accurate and personalized evaluation of optic nerve function through innovative signal processing technology, machine learning algorithm and multimodal data fusion strategy.

[0006] The present invention provides an intelligent evaluation method for optic nerve function based on visual evoked potential, comprising:

[0007] The acquisition steps include:

[0008] Obtain the subject's EEG signals, eye movement trajectories, and auditory response signals;

[0009] Acquiring visual stimulation parameters, wherein the visual stimulation parameters include stimulation frequency, stimulation intensity and stimulation duration;

[0010] Processing steps include:

[0011] Based on the EEG signal, the eye movement trajectory and the auditory response signal, a signal preprocessing operation is performed to obtain a preprocessed multimodal physiological signal;

[0012] Based on the preprocessed multimodal physiological signals, an improved spatiotemporal convolutional neural network is used to extract features to obtain optic nerve function features;

[0013] Establishing an optic nerve conduction pathway model according to the optic nerve functional characteristics and the visual stimulation parameters;

[0014] Based on the optic nerve conduction pathway model, evaluating the functional state of the optic nerve to obtain an optic nerve function evaluation result;

[0015] Output steps include:

[0016] generating an assessment report including the optic nerve function assessment results;

[0017] Based on the optic nerve function assessment results, the optic nerve function change trend is predicted and a prediction report is generated.

[0018] Preferably, the signal preprocessing operation specifically includes:

[0019] Performing high-pass filtering on the EEG signal to remove power frequency interference and baseline drift;

[0020] Performing fast Fourier transform on the EEG signal to obtain frequency domain features;

[0021] Performing wavelet transform on the EEG signal to obtain time-frequency characteristics;

[0022] Multimodal data fusion is performed on the EEG signal, eye movement trajectory and auditory response signal.

[0023] Preferably, the improved spatiotemporal convolutional neural network comprises:

[0024] Multiple convolutional layers to extract spatial features;

[0025] Multiple pooling layers to reduce feature dimensionality;

[0026] Multiple fully connected layers to capture temporal features;

[0027] A Softmax classification layer is used to output the classification results.

[0028] Preferably, the method further comprises an adaptive stimulation step:

[0029] Dynamically adjusting the visual stimulation parameters based on the optic nerve function assessment result;

[0030] generating a new visual stimulation sequence according to the adjusted visual stimulation parameters;

[0031] The new visual stimulus sequence is applied to the subject and the acquisition and processing steps are repeated.

[0032] Preferably, the process of establishing the optic nerve conduction pathway model includes:

[0033] Based on the functional characteristics of the optic nerve, an end-to-end model from visual stimulation to the state of the optic nerve conduction pathway is constructed;

[0034] Using a transfer learning method, initializing parameters of the end-to-end model using a pre-trained model;

[0035] The end-to-end model is fine-tuned using the current subject's data.

[0036] Preferably, the optic nerve function assessment results include:

[0037] The degree of damage to the optic nerve conduction pathway;

[0038] The changing trend of optic nerve function;

[0039] Risk assessment based on the extent and trend of damage described.

[0040] Preferably, an objective quantitative evaluation step is also included:

[0041] Calculating multiple objective evaluation indicators based on the optic nerve function evaluation results;

[0042] Comparing the objective evaluation index with a preset normal value range;

[0043] Based on the comparison results, the degree of abnormality of optic nerve function is determined.

[0044] Preferably, the objective evaluation indicators include:

[0045] mean evoked latency;

[0046] Average evoked peak value;

[0047] Optic nerve conduction velocity;

[0048] Morphological characteristics of visual evoked potential waveforms.

[0049] Preferably, a multi-level evaluation step is also included:

[0050] Generate an evaluation report containing subjective scores;

[0051] Generate an evaluation report containing objective quantitative indicators;

[0052] Generate an assessment report containing risk profiles;

[0053] The subjective scores, objective quantitative indicators and risk predictions are comprehensively analyzed to obtain the final optic nerve function assessment conclusion.

[0054] The intelligent evaluation system of optic nerve function based on visual evoked potential for executing the method comprises:

[0055] Data acquisition module for:

[0056] Collect the subjects' EEG signals, eye movement trajectories and auditory response signals;

[0057] Obtain visual stimulus parameters;

[0058] Preprocessing module for:

[0059] Preprocessing the EEG signal, eye movement trajectory and auditory response signal to obtain a preprocessed multimodal physiological signal;

[0060] Feature extraction module for:

[0061] Based on the preprocessed multimodal physiological signals, an improved spatiotemporal convolutional neural network is used to extract optic nerve functional features;

[0062] Model building modules for:

[0063] Establishing an optic nerve conduction pathway model based on the optic nerve functional characteristics and the visual stimulation parameters;

[0064] Evaluation modules for:

[0065] Based on the optic nerve conduction pathway model, evaluating the functional state of the optic nerve to obtain an optic nerve function evaluation result;

[0066] Output modules for:

[0067] generating an assessment report including the optic nerve function assessment results;

[0068] Based on the optic nerve function assessment results, predict the trend of optic nerve function changes and generate a prediction report;

[0069] Adaptive stimulation module for:

[0070] Dynamically adjusting the visual stimulation parameters based on the optic nerve function assessment result;

[0071] New visual stimulus sequences were generated and applied to the subjects.

[0072] The beneficial effects of the present invention are mainly reflected in the following aspects:

[0073] First, the present invention uses an improved spatiotemporal convolutional neural network to extract the features of VEP signals, greatly improving the efficiency and accuracy of feature extraction. The network can simultaneously capture the temporal and spatial features of VEP signals, thereby more comprehensively reflecting the functional state of the optic nerve. Compared with traditional manual feature extraction methods, this deep learning method can automatically learn and adapt to different types of VEP signal features, significantly improving the accuracy and generalization ability of the assessment.

[0074] Secondly, the present invention introduces an adaptive stimulation strategy that can dynamically adjust visual stimulation parameters according to the real-time response of the subject. This closed-loop feedback mechanism greatly improves the personalization and accuracy of the evaluation. By continuously optimizing the stimulation scheme, the system can obtain the most effective VEP response in the shortest time, which not only improves the evaluation efficiency but also reduces the discomfort of the subject.

[0075] Furthermore, the present invention uses multimodal data fusion and comprehensive utilization of multiple physiological signals such as EEG, eye movement and auditory response, which greatly enriches the information dimension of the assessment. This collaborative analysis of multi-source information not only improves the accuracy of the assessment, but also reveals complex neural function states that are difficult to reflect with a single VEP signal.

[0076] In addition, the present invention constructs an end-to-end optic nerve conduction pathway model, realizing the full simulation from visual stimulation to neural response. This holistic modeling method not only provides more accurate functional evaluation results, but also provides a new research tool for understanding the mechanism of visual information processing.

[0077] Finally, the method of the present invention can not only evaluate the current state of optic nerve function, but also predict its future change trend. This prediction function is of great significance for the early diagnosis and prevention of diseases, and can help doctors formulate more reasonable treatment plans and follow-up plans.

[0078] In summary, the intelligent optic nerve function assessment method and system provided by the present invention achieves the unity of high efficiency, accuracy, personalization and predictability of assessment through the organic combination of multiple innovative technologies. This not only greatly enhances the value of VEP technology in clinical applications, but also opens up a new path for the study of optic nerve-related diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 The figure is a flow chart of the method of the present invention.

[0080] Figure 2 It is a logic block diagram of the data acquisition module of the present invention.

[0081] Figure 3It is a logic block diagram of the preprocessing module of the present invention.

[0082] Figure 4 It is a logic block diagram of the feature extraction module of the present invention.

[0083] Figure 5 A logical block diagram of the model building module of the present invention.

[0084] Figure 6 It is a logic block diagram of the evaluation module of the present invention.

[0085] Figure 7 It is a logic block diagram of the adaptive stimulation module of the present invention.

[0086] Figure 8 It is a logic block diagram of the output module of the present invention. DETAILED DESCRIPTION

[0087] Please refer to Figure 1-8 The present invention provides a method and system for intelligently evaluating optic nerve function based on visual evoked potential. The method realizes intelligent evaluation of optic nerve function by collecting and analyzing visual evoked potential signals and combining advanced signal processing technology and machine learning algorithms.

[0088] First, the method includes an acquisition step, a processing step, and an output step. In the acquisition step, the system acquires the subject's EEG signal, eye movement trajectory, and auditory response signal, and simultaneously acquires visual stimulation parameters. These visual stimulation parameters include stimulation frequency, stimulation intensity, and stimulation duration. Preferably, the stimulation frequency can be set within the range of 1-100 Hz, and the stimulation intensity can be set within the range of 0.1-10 cd / m 2 The stimulation duration can be set to 50-500ms. These parameters are selected based on a large amount of clinical experimental data to ensure that visual potential signals can be effectively induced.

[0089] In the processing step, the method of the present invention first preprocesses the acquired signal. Specifically, the system performs a signal preprocessing operation to obtain a preprocessed multimodal physiological signal. This step is crucial because it can significantly improve the accuracy of subsequent analysis. For example, high-pass filtering of the EEG signal can effectively remove 50Hz or 60Hz power frequency interference and improve the signal-to-noise ratio of the signal.

[0090] Next, the system uses an improved spatiotemporal convolutional neural network to extract features based on the preprocessed multimodal physiological signals to obtain optic nerve function features. Here, the improved spatiotemporal convolutional neural network is a key innovation of the present invention. The network structure is as follows:

[0091] Output = Softmax(FC n (…FC2(FC1(Poolm (…Pool2(Pool1(Conv k (…Conv2(Conv1(Input))))))…),

[0092] Among them, Conv i represents the i-th convolutional layer, Pool j represents the jth pooling layer, FC l represents the lth fully connected layer. This network structure can effectively capture the spatiotemporal characteristics of EEG signals, thereby improving the accuracy of feature extraction.

[0093] After feature extraction, the system builds an optic nerve conduction pathway model based on the optic nerve functional characteristics and visual stimulation parameters. This model is an end-to-end model that can simulate the entire process from visual stimulation to the state of the optic nerve conduction pathway. The mathematical expression of the model is as follows:

[0094] S=f(V,P),

[0095] Among them, S represents the state of the optic nerve conduction pathway, V represents the functional characteristics of the optic nerve, P represents the visual stimulation parameters, and f represents the model function.

[0096] Based on the established optic nerve conduction pathway model, the system evaluates the optic nerve function status and obtains the optic nerve function evaluation results. This evaluation process uses a machine learning algorithm that can automatically learn and adapt to the characteristics of different subjects.

[0097] In the output step, the system generates an evaluation report containing the optic nerve function evaluation results. This report not only contains the current evaluation results, but also includes the predicted trend of optic nerve function changes based on the evaluation results. This prediction function is of great significance for the early diagnosis and prevention of optic nerve-related diseases.

[0098] The method of the present invention also includes an innovative signal preprocessing operation. Specifically, the system performs high-pass filtering on the EEG signal to remove power frequency interference and baseline drift. The cutoff frequency of the high-pass filter is usually set in the range of 0.1-1 Hz, which can effectively remove low-frequency noise without losing useful signals.

[0099] Next, the system performs a fast Fourier transform on the EEG signal to obtain the frequency domain features. The mathematical expression of the fast Fourier transform is as follows:

[0100]

[0101] Among them, x(n) is the time domain signal, X(k) is the frequency domain signal, N is the signal length, k = 0, 1, ..., N-1. In addition, the system also performs wavelet transform on the EEG signal to obtain the time-frequency characteristics. The mathematical expression of wavelet transform is as follows:

[0102]

[0103] Among them, x(t) is the input signal, ψ(t) is the wavelet function, a is the scale parameter, and b is the translation parameter. Finally, the system performs multimodal data fusion on EEG signals, eye movement trajectories, and auditory response signals. This multimodal fusion method can comprehensively utilize different types of physiological signals to provide more comprehensive information on optic nerve function.

[0104] Finally, the system performs multimodal data fusion on EEG signals, eye movement trajectories, and auditory response signals. This multimodal fusion method can comprehensively utilize different types of physiological signals to provide more comprehensive information on optic nerve function.

[0105] The improved spatiotemporal convolutional neural network of the present invention includes multiple convolutional layers, pooling layers, fully connected layers and a Softmax classification layer. The convolutional layer is used to extract spatial features, the pooling layer is used to reduce feature dimensions, the fully connected layer is used to capture temporal features, and the Softmax classification layer is used to output classification results.

[0106] The mathematical expression of the convolutional layer is as follows:

[0107] y j =f(Σ i x i *k ij +b j ),

[0108] Among them, x i is the input feature map, k ij is the convolution kernel, b j is the bias term and f is the activation function.

[0109] The pooling layer usually uses maximum pooling or average pooling, and their mathematical expressions are as follows:

[0110] Max Pooling:

[0111]

[0112] Average Pooling:

[0113]

[0114] Among them, R represents the pooling area and |R| represents the area size.

[0115] The mathematical expression of the fully connected layer is as follows:

[0116] y=f(Wx+b),

[0117] Where W is the weight matrix, x is the input vector, b is the bias vector, and f is the activation function.

[0118] The mathematical expression of the Softmax classification layer is as follows:

[0119]

[0120] Among them, z i is the score of the ith category, p i is the probability of the ith category.

[0121] Through this network structure, the method of the present invention can effectively capture the spatiotemporal characteristics of visual evoked potential signals, thereby improving the accuracy of optic nerve function assessment. At the same time, the network structure has good interpretability, which is conducive to doctors understanding the reasons for the assessment results.

[0122] In summary, the intelligent evaluation method and system of optic nerve function based on visual evoked potential provided by the present invention realizes accurate evaluation and prediction of optic nerve function through innovative signal processing technology and machine learning algorithm. This method can not only improve the diagnostic accuracy of optic nerve-related diseases, but also provide an important basis for early prevention, and has important clinical application value. The method of the present invention also includes an innovative adaptive stimulation step. In this step, the system dynamically adjusts the visual stimulation parameters based on the optic nerve function evaluation results. This adaptive mechanism can optimize the stimulation scheme according to the individual differences of each subject, so as to obtain more accurate evaluation results.

[0123] Specifically, the system first adjusts the visual stimulation parameters based on the initial evaluation results. For example, if the initial evaluation shows that the subject has a weak response to stimulation of a certain frequency, the system may increase the intensity or duration of stimulation of that frequency. Preferably, the adjustment range of the stimulation frequency can be ±20% of the initial frequency, the adjustment range of the stimulation intensity can be ±30% of the initial intensity, and the adjustment range of the stimulation duration can be ±25% of the initial duration. These adjustment ranges are determined based on a large amount of clinical experimental data, which can ensure the stimulation effect while avoiding discomfort to the subject.

[0124] After adjusting the parameters, the system generates a new visual stimulus sequence based on the new visual stimulus parameters. This new stimulus sequence is applied to the subject, and then the system repeats the acquisition step and the processing step. Through this iterative process, the system can continuously optimize the stimulus scheme and ultimately obtain the best evaluation results.

[0125] In a preferred embodiment of the present invention, the process of establishing the optic nerve conduction pathway model includes several key steps. First, based on the functional characteristics of the optic nerve, the system constructs an end-to-end model from visual stimulation to the state of the optic nerve conduction pathway. This model can simulate the entire visual information processing process, from the retina receiving light signals, to the optic nerve conduction, and then to the signal processing of the visual cortex of the brain.

[0126] In the process of model construction, this method adopts an innovative transfer learning method. Specifically, the system uses a pre-trained model to initialize the parameters of the end-to-end model. This pre-trained model is trained on a large-scale visual evoked potential dataset and contains rich prior knowledge. Through transfer learning, the system can quickly build an accurate model based on limited current subject data.

[0127] Preferably, the structure of the pre-trained model can be a multi-layer perceptron or a recursive neural network. For example, a typical multi-layer perceptron model can include 3-5 hidden layers, each layer including 50-200 neurons. The recursive neural network model can use long short-term memory (LSTM) units, usually including 2-3 layers of LS TM layers, each layer including 100-300 LSTM units. These model structure parameters are based on a large number of experiments and can achieve a good balance between model complexity and computational efficiency.

[0128] After transfer learning, the system will use the current subject's data to fine-tune the end-to-end model. The fine-tuning process usually uses the stochastic gradient descent algorithm, the learning rate can be set to 0.001-0.01, and the batch size can be set to 16-64. The selection of these hyperparameters can ensure that the model converges quickly while avoiding overfitting.

[0129] In the method of the present invention, the optic nerve function assessment results include multiple important indicators. The first is the degree of damage to the optic nerve conduction pathway. This indicator is usually expressed as a score of 0-100, where 0 represents no damage and 100 represents severe damage. Preferably, the degree of damage can be further subdivided into mild (0-30), moderate (31-60) and severe (61-100). This subdivision method can provide a more accurate reference for clinical diagnosis.

[0130] Secondly, the evaluation results also include the trend of changes in optic nerve function. This trend is usually represented by a vector, for example [+1, -1, +2] may mean that in the next three months, optic nerve function is expected to improve slightly, then decline slightly, and finally improve significantly. This trend prediction is of great guiding significance for formulating long-term treatment plans.

[0131] Finally, based on the degree of injury and the trend of change, the system will generate a risk assessment report. This report usually contains three risk levels: low, medium, and high, each with detailed explanations and suggestions. For example, high risk may mean that further medical examinations or treatment should be started immediately, while low risk may only require regular monitoring.

[0132] The method of the present invention also includes an innovative objective quantitative evaluation step. In this step, the system first calculates multiple objective evaluation indicators based on the optic nerve function evaluation results. These indicators are strictly verified and can objectively reflect various aspects of optic nerve function.

[0133] Next, the system compares these objective evaluation indicators with the preset normal range. This normal range is based on data statistics of a large number of healthy people. Preferably, the normal range can be subdivided according to factors such as age and gender to improve the accuracy of the comparison. For example, for a 20-30 year old male, the normal range of a certain indicator may be 80-120, while for a 60-70 year old female, this range may be 70-110.

[0134] Finally, based on the comparison results, the system determines the degree of abnormality of the optic nerve function. The degree of abnormality is usually divided into three levels: mild, moderate and severe. Preferably, if a certain indicator exceeds the normal range but does not exceed 20%, it can be determined as a mild abnormality; if it exceeds 20%-50%, it can be determined as a moderate abnormality; if it exceeds more than 50%, it can be determined as a severe abnormality. This grading method can provide doctors with a clear diagnostic reference.

[0135] Through this objective quantitative evaluation, the method of the present invention can provide more accurate and reliable optic nerve function evaluation results, effectively avoiding the deviations and inconsistencies that may be caused by subjective evaluation. This is of great significance for improving diagnostic accuracy and formulating personalized treatment plans. In the method of the present invention, the objective evaluation indicators include multiple key parameters, which can comprehensively reflect all aspects of optic nerve function. The first is the average induced latency, which reflects the time required for visual stimuli to reach the visual cortex from the retina. Preferably, the normal range of the average induced latency is usually between 80-120 milliseconds. If the latency is significantly prolonged, it may mean that the optic nerve conduction velocity is slowed, which is an important indicator of impaired optic nerve function.

[0136] The second is the average evoked peak value, which reflects the strength of the visual evoked potential. Under normal circumstances, the average evoked peak value should be between 5-15 microvolts. A decrease in the peak value may indicate impaired optic nerve function or decreased visual cortical responsiveness. It is worth noting that the absolute size of the peak value may vary due to individual differences, so when evaluating, it is more important to focus on the changing trend of the same individual at different time points.

[0137] Optic nerve conduction velocity is another important objective assessment indicator. This indicator can be obtained by calculating the distance from the retina to the visual cortex of the visual stimulus divided by the latency. Normal optic nerve conduction velocity is usually between 40-60 m / s. A significant decrease in conduction velocity may mean damage to the optic nerve myelin sheath or axonal degeneration.

[0138] Finally, the morphological characteristics of the visual evoked potential waveform are also an indicator that cannot be ignored. A normal visual evoked potential waveform usually includes three main components: N75, P100, and N135. Among them, P100 is the most stable and important component, and changes in its latency and amplitude are of great significance for the diagnosis of optic neuropathy. For example, if the P100 latency is prolonged by more than 10 milliseconds, or the difference in P100 latency between the left and right eyes exceeds 6 milliseconds, it may indicate abnormal optic nerve function.

[0139] In a preferred embodiment of the present invention, a multi-level evaluation step is also included. This step is intended to provide a more comprehensive and in-depth evaluation result of optic nerve function. First, the system generates an evaluation report containing a subjective score. This subjective score is usually given by an experienced clinician based on the patient's symptom description, vision test results and other information. Although it is highly subjective, this score can reflect some clinical manifestations that are difficult to quantify and is of great value for a comprehensive assessment of the patient's condition.

[0140] Next, the system will generate an evaluation report containing objective quantitative indicators. This report includes the various objective evaluation indicators mentioned above, such as average evoked latency, average evoked peak, optic nerve conduction velocity, etc. These indicators have clear values ​​and reference ranges, which can provide objective and comparable evaluation results.

[0141] The system then generates an assessment report containing risk predictions. This report is based on the current assessment results and combines machine learning models to predict the possible trend of future changes in the patient's optic nerve function. For example, the report may state: "Based on the current assessment results, the patient's risk of further deterioration of optic nerve function in the next 6 months is 30%." This predictive assessment is of great guiding significance for formulating preventive treatment plans.

[0142] Finally, the system will conduct a comprehensive analysis of subjective scores, objective quantitative indicators and risk prediction to draw the final conclusion of optic nerve function assessment. This comprehensive analysis process adopts a weighted average method, in which objective quantitative indicators have the highest weight, usually accounting for 50-60%, risk prediction accounts for 30-40%, and subjective scores account for 10-20%. This weight distribution can fully consider clinical experience and future risks while ensuring the objectivity of the assessment.

[0143] The present invention also provides an intelligent evaluation system for optic nerve function based on visual evoked potential, which includes multiple functional modules, each of which has its specific function and role.

[0144] The data acquisition module 1 is used to collect the subject's EEG signals, eye movement trajectories and auditory response signals, and obtain visual stimulation parameters at the same time. This module usually includes an electroencephalograph, an eye tracker and an auditory stimulation device. Preferably, the sampling rate of the electroencephalograph should be no less than 1000Hz to ensure that high-frequency components can be captured; the sampling rate of the eye tracker should be no less than 250Hz to accurately record rapid eye movements; the auditory stimulation device should be able to produce pure tone stimulation in the range of 20-20000Hz.

[0145] Preprocessing module 2 is responsible for preprocessing the collected signals to obtain preprocessed multimodal physiological signals. This module contains multiple subunits, such as filtering unit, artifact removal unit and signal enhancement unit. The filtering unit is mainly used to remove 50Hz or 60Hz power frequency interference and other high-frequency noise; the artifact removal unit is used to identify and remove artifacts such as blinking and electromyography; the signal enhancement unit improves the signal-to-noise ratio of the signal through methods such as wavelet transform.

[0146] Feature extraction module 3 uses an improved spatiotemporal convolutional neural network to extract optic nerve functional features based on preprocessed multimodal physiological signals. This module is one of the core parts of the system. It uses deep learning technology to automatically learn and extract complex spatiotemporal features. The network structure includes multiple convolutional layers, pooling layers, and fully connected layers. The specific parameters can be optimized and determined by methods such as cross-validation.

[0147] Model building module 4 is responsible for building the optic nerve conduction pathway model based on the functional characteristics of the optic nerve and the visual stimulation parameters. This module adopts an end-to-end modeling method and can directly learn the state of the optic nerve conduction pathway from the original data. The model training process adopts transfer learning technology, first pre-training on a large-scale data set, and then fine-tuning on specific patient data. This method can effectively solve the problem of insufficient sample size.

[0148] The evaluation module 5 evaluates the functional state of the optic nerve based on the optic nerve conduction pathway model and obtains the optic nerve function evaluation result. This module contains multiple evaluation subunits, each corresponding to different evaluation indicators, such as the damage degree evaluation unit, the functional change trend evaluation unit, etc. Each subunit adopts a specific algorithm. For example, the damage degree evaluation may adopt the support vector machine algorithm, while the functional change trend evaluation may adopt the long short-term memory network.

[0149] Output module 6 is responsible for generating an evaluation report containing the evaluation results of optic nerve function, and predicting the trend of optic nerve function changes based on the evaluation results to generate a prediction report. This module is not just a simple data display, but also includes advanced functions such as data visualization and natural language generation. For example, it can generate easy-to-understand charts to display the evaluation results, and generate easy-to-understand text descriptions to explain these results.

[0150] Finally, the adaptive stimulation module 7 dynamically adjusts the visual stimulation parameters based on the optic nerve function evaluation results, generates a new visual stimulation sequence and applies it to the subject. This module implements the closed-loop feedback function of the system, and can optimize the stimulation scheme according to the real-time evaluation results to obtain more accurate evaluation results. Preferably, the module adopts a reinforcement learning algorithm to achieve adaptive adjustment of parameters, and can gradually find the optimal stimulation parameters in multiple iterations.

[0151] Through the coordinated work of these functional modules, the system of the present invention can achieve efficient and accurate intelligent evaluation of optic nerve function, providing strong support for clinical diagnosis and treatment.

[0152] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, replacement, and improvement made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent evaluation method for optic nerve function based on visual evoked potential, characterized in that: include: The acquisition steps include: Obtain the subject's EEG signals, eye movement trajectories, and auditory response signals; Acquiring visual stimulation parameters, wherein the visual stimulation parameters include stimulation frequency, stimulation intensity and stimulation duration; Processing steps include: Based on the EEG signal, the eye movement trajectory and the auditory response signal, a signal preprocessing operation is performed to obtain a preprocessed multimodal physiological signal; Based on the preprocessed multimodal physiological signals, an improved spatiotemporal convolutional neural network is used to extract features to obtain optic nerve function features; Establishing an optic nerve conduction pathway model according to the optic nerve functional characteristics and the visual stimulation parameters; Based on the optic nerve conduction pathway model, evaluating the functional state of the optic nerve to obtain an optic nerve function evaluation result; Output steps include: generating an assessment report including the optic nerve function assessment results; Based on the optic nerve function assessment results, the optic nerve function change trend is predicted and a prediction report is generated.

2. The method according to claim 1, characterized in that: The signal preprocessing operation specifically includes: Performing high-pass filtering on the EEG signal to remove power frequency interference and baseline drift; Performing fast Fourier transform on the EEG signal to obtain frequency domain features; Performing wavelet transform on the EEG signal to obtain time-frequency characteristics; Multimodal data fusion is performed on the EEG signal, eye movement trajectory and auditory response signal.

3. The method according to claim 1, characterized in that The improved spatiotemporal convolutional neural network includes: Multiple convolutional layers to extract spatial features; Multiple pooling layers to reduce feature dimensionality; Multiple fully connected layers to capture temporal features; A Softmax classification layer is used to output the classification results.

4. The method according to claim 1, characterized in that Also included is an adaptive stimulation step: Dynamically adjusting the visual stimulation parameters based on the optic nerve function assessment result; generating a new visual stimulation sequence according to the adjusted visual stimulation parameters; The new visual stimulus sequence is applied to the subject and the acquisition and processing steps are repeated.

5. The method according to claim 1, characterized in that The process of establishing the optic nerve conduction pathway model includes: Based on the functional characteristics of the optic nerve, an end-to-end model from visual stimulation to the state of the optic nerve conduction pathway is constructed; Using a transfer learning method, initializing parameters of the end-to-end model using a pre-trained model; The end-to-end model is fine-tuned using the current subject's data.

6. The method according to claim 1, characterized in that The optic nerve function assessment results include: The degree of damage to the optic nerve conduction pathway; Trends in optic nerve function; Risk assessment based on the extent and trend of damage described.

7. The method according to claim 1, characterized in that It also includes objective quantitative assessment steps: Calculating multiple objective evaluation indicators based on the optic nerve function evaluation results; Comparing the objective evaluation index with a preset normal value range; Based on the comparison results, the degree of abnormality of optic nerve function is determined.

8. The method according to claim 7, characterized in that The objective evaluation indicators include: mean evoked latency; Average evoked peak value; Optic nerve conduction velocity; Morphological characteristics of visual evoked potential waveforms.

9. The method according to claim 1, characterized in that: It also includes a multi-level assessment step: Generate an evaluation report containing subjective scores; Generate an evaluation report containing objective quantitative indicators; Generate an assessment report containing risk profiles; The subjective scores, objective quantitative indicators and risk predictions are comprehensively analyzed to obtain the final optic nerve function assessment conclusion.

10. An intelligent optic nerve function assessment system based on visual evoked potentials that implements the method according to any one of claims 1 to 9, characterized in that: include: Data acquisition module for: Collect the subjects' EEG signals, eye movement trajectories and auditory response signals; Obtain visual stimulus parameters; Preprocessing module for: Preprocessing the EEG signal, eye movement trajectory and auditory response signal to obtain a preprocessed multimodal physiological signal; Feature extraction module for: Based on the preprocessed multimodal physiological signals, an improved spatiotemporal convolutional neural network is used to extract optic nerve functional features; Model building modules for: Establishing an optic nerve conduction pathway model based on the optic nerve functional characteristics and the visual stimulation parameters; Evaluation modules for: Based on the optic nerve conduction pathway model, evaluating the functional state of the optic nerve to obtain an optic nerve function evaluation result; Output modules for: generating an assessment report including the optic nerve function assessment results; Based on the optic nerve function assessment results, predict the trend of optic nerve function changes and generate a prediction report; Adaptive stimulation module for: Dynamically adjusting the visual stimulation parameters based on the optic nerve function assessment result; New visual stimulus sequences were generated and applied to the subjects.

Images